9:44 a.m.
Here's my notes from the RDO community meetup at the OpenStack Summit
last week. If you have major changes/additions/whatever that you'd like
to make, you can do that at
https://etherpad.openstack.org/p/rdo_community_meetup_summit and I can
re-post it later on. Please forgive any omissions or errors - it was
very hard to hear a lot of what was said.
--------------------
Notes from RDO community meetup at the OpenStack Summit in paris
Please forgive errors - notes taken very quickly, and it was very
difficult to hear.
In attendance, perhaps 60 people, but most (50?) were from Red Hat. The
goal of this meetup was to engage folks from outside of Red Hat. These
are some (not all) of the non-RedHat people in attendance:
Tim Bell - CERN
Francesco - Public Health England
Arif Ali - OCF plc
Jan Ivan - Universities from Norway - Norcom
VNFN (?)
Comments from these folks on what we're doing right/wrong
Tim: The deployability of what we're producing. Prerequisites handled.
Works right away with minimal fiddling.
First activity of a new CERN IT employee is to spin up a new OpenStack
cloud using RDO. (Same at Norcom.)
Would like to have EL6 packages - not of everything, but certain pieces.
Would like to see more automated handling (reports, tickets) coming out
of the Jenkins instance running on TryStack. This is mostly hidden from
the community right now, and we need to do a beter job of exposing that.
(ACTION: Dan Radez, Lars Kellog-Stedman, Rich, follow up on this.)
What tests are we running there? This needs to be documented (or, at
least, what we already have needs to be better publicized.
Advertise what scenarios we're covering, and make it easy for people to
contribute new scenarios that we can run.
Publish links to the Jenkins instance and results. (ACTION: Rich needs
to follow up with Dan on this.)
Better document what bugs go where. Almost everything goes to upstream.
Packstack has upstream ticket tracking. Only things directly related to
packaging should come to RDO bugzilla.
Document when, how often we trigger a packaging run. What's the "barrier
to entry"? Could we enable other folks to package the bits they care about?
Need to upstream rdopkg, and have more information about our Delorian
effort. It sounds like Delorian is a good place to move a lot of our
effort, so that packaging activity is happening on trunk, and is all in
the open.
Derek Higgins talked about Delorian. It's in the openstack-packaging
repo on Github. Delorian builds trunk packages.
Can we add .deb packages to this project? Derek says that there's no
reason that we wouldn't be able to do this, if someone from the
community steps up to do this.
Need doc on RDO site on what/where the Delorian project is, and how one
might get involved. (ACTION: Rich)
Hugh asked that we devote some time to discussing how to converge
various deployment methods. A matrix of all of the various methods, and
when each is useful, would be hugely helpful. Would also facilitate
understanding where we can start to combine and converge into a single
project.
Arif asked about the Foreman installer for EL7. Existing docs are for
EL6 and he's been working on get things documented for EL7, but there's
some missing bits.
Alan talked about the process of getting involved as a contributor to
RDO (See
https://www.redhat.com/archives/rdo-list/2014-November/msg00018.html )
Tim expressed concern with having to become a Fedora contributor, since
just signing the OpenStack CLA was an 8 month process at CERN. Kashyap
says that it's not actually a CLA, and that he's going to look into
details on this, and get them to Tim and/or to rdo-list (ACTION: Kashyap?)
The CentOS Cloud SIG was briefly mentioned, Karsten Wade suggested that
we get Karanbir Singh, Alan Pevec, Kashyap Chamarthy, and other
intereste people into a hangout to discuss the way forward on this, now
that the packaging picture is more widely understood. (Shoud start
another thread on this.)
Can do better: Clearly communicate, well in advance, what the gaps are
in what we (Red Hat) are able to do, do that the community can step up
to fill those gaps. The "Nothing for EL6" announcement took a lot of
people by surprise, and people would have done things very differently
if we had communicated this earlier.
Hugh broght up again - how do we get to one uniified deployment/installer?
Derek: Important to settle on one set of Puppet modules, from the many
overlapping ones that we're currently using.
A glossary/matrix of installers needs to be on the RDO site, with
capabilities and scenarios. (ACTION: Rich, Hugh, Perry, Arif collaborate
on this?)
Some that were mentioned were:
Spinal Stack (Emilien Macchi - eNovance)
QuickStack
Instack (James Slagle)
Packstack
Things began to wrap up at this point. If you have more notes or
commentary, please add to this conversation, and/or spin off a new
thread on major topics.
--
Rich Bowen - rbowen(a)rcbowen.com - @rbowen
http://apachecon.com/ - @apachecon